MAE PhD Defense – Opeoluwa Owoyele
December 20 @ 9:00 am - 11:00 am
TITLE: Accelerating the Simulation of Chemically Reacting Turbulent Flows via Machine Learning Techniques
ADVISOR: Dr. Tarek Echekki
DATE & TIME: Wednesday, December 20, 2017 at 9 AM
LOCATION: EB3 – 3235
Turbulent reacting flows are one of the most complex class of engineering problems, because in addition to the chaotic process of turbulence, they involve stiff and highly nonlinear chemical kinetics. The presence of a large number of species as well as a large range of scales in space and time make direct numerical simulations (DNS) of these processes computationally prohibitive. In this study, the application of machine learning to reduce computational costs of the simulation of this class of flows is explored. Principal components analysis, combined with artificial neural networks (PCA-ANN) is applied to the temporal integration of stiff thermochemical equations in time in the context of DNS. In addition to being used for stand-alone models, machine learning can also be applied in order to enhance existing models. Thus, this work also investigates the use of the tabulated, multidimensional unsteady flamelet model and artificial neural networks (TFM-ANN) for lifted diesel spray flame applications. Overall, the deep learning tools and algorithms developed and investigated in this study can be extended to other combustion problems and models.
Opeoluwa (Ope) obtained his Bachelors of Engineering in 2011 from University of Ilorin, Nigeria, where he graduated top of his class. Upon graduation, he undertook the National Youth Service for one year, where he taught Mathematics, Physics and Chemistry to high-school students in a village that didn’t have access to good education. In January 2013, he started his masters in Mechanical Engineering at North Carolina State University, where he worked with Dr. Brendan O’Connor on developing computational and theoretical models involving heat transfer in a thick film thermoelectric device architecture for wearable device applications. He started his PhD in the fall of 2014 in the same university under Prof. Tarek Echekki, working on applying machine learning to turbulent combustion simulations. His research interests include turbulent reacting flows, multiphase flows and machine learning.